Event conditioning
Noise filtering, hot-pixel rejection and ego-motion compensation applied to the raw event stream before anything downstream sees it.
Capabilities
What we deliver
Four lines of work, all pointed at the same problem: closing the sensing and decision loop fast enough to matter in a terminal engagement against small, fast, low-signature air targets.
01 — Sensor Payloads
Event-based seeker heads
Integrated sensor assemblies built around event-based (neuromorphic) imagers, optionally paired with a conventional frame or GigE Vision channel for classification and operator confidence. Delivered as a mechanical, electrical and thermal package sized to the host — not as a lab bench.
- Event-only and dual-modality (event + frame) configurations
- Optical design, athermalization and boresight retention
- On-head processing: FPGA front end with embedded SoC for track logic
- Environmental and captive-carry qualification against program requirements
- Defined electrical and data interfaces for airframe or mount integration
Why event sensing for a seeker
An event pixel reports a log-intensity change the moment it happens, independently of its neighbors. There is no exposure window, so there is no motion blur; there is no global gain, so a target does not disappear into a bright sky; and there is no redundant background data, so the processing load scales with scene change rather than with resolution.
That combination is what makes microsecond-class temporal resolution affordable inside a mass, power and cost envelope an expendable effector can actually carry.
02 — GNC
Guidance, navigation & fire control
Turning a track into a command. We build the estimation and guidance layer that sits between the seeker and the effector, and we build it to a latency budget rather than to a benchmark score.
- Extended Kalman filtering and visual-inertial estimation for agile targets
- Track initiation, association and custody through occlusion and clutter
- Proportional navigation and terminal guidance law implementation
- Aided-aim and lead-angle solutions for direct-fire weapons
- Cue, handoff and track-quality reporting to an external fire control system
| Design target | Deterministic detect-to-command latency in the single-digit millisecond class |
|---|---|
| Estimation | EKF / VIO with IMU fusion; sensor-agnostic front end |
| Compute | FPGA fabric for event preprocessing; embedded SoC for track and guidance logic |
| Interfaces | Program-defined; Ethernet, serial and discrete options |
| Verification | Hardware-in-the-loop, captive carry, and live flight test on private range |
03 — Software
Edge autonomy stack
The sensor is the easy half. The stack that turns a sparse asynchronous event stream into a trustworthy track is the part that takes years.
Detection & clustering
Spatio-temporal clustering that finds coherent moving structure without ever reconstructing a frame — the step that keeps the compute budget honest.
Track & classify
Multi-target track formation with lightweight on-edge classification to separate threat air vehicles from birds, debris and background motion.
04 — Services
Contract R&D and SBIR partnership
Much of this capability reaches the field through other people's programs. We take on the sensing and autonomy scope inside a larger effort, as a subcontractor to a prime or as principal investigator on our own award.
- SBIR/STTR Phase I, II and III execution and transition planning
- Sensor trade studies and feasibility analysis against a specific engagement
- Prototype design, build and test on a program schedule
- Algorithm development and porting onto customer-specified edge hardware
- Data collection campaigns on our own range
Facility and test. 7,000 square feet of in-house design, fabrication and test space plus a 13-acre private outdoor range, so design, build and field test iterate without scheduling an external site.
Export control. This site describes capability only. Controlled technical data is exchanged under appropriate agreement and authorization, not published here.